Comparative Evaluation of Five Obesity Indices for Cardiovascular and All-Cause Mortality Prediction in Prediabetes and Diabetes Population: A Prospective Cohort.

BACKGROUND: In the US population, the association and predictive performance of obesity indices -- Body Mass Index (BMI), Body Shape Index (ABSI), Body Roundness Index (BRI), Waist Circumference (WC), and Waist-to-Height Ratio (WHtR) with mortality outcomes Cardiovascular Mortality (CVM) and All-Cause Mortality (ACM)) in prediabetes and diabetes (DM) populations have not been systematically evaluated. METHODS: Utilizing NHANES data (prediabetes: n = 10,988, DM: n = 5,868), we evaluated these associations and predictive ability using Kaplan-Meier curves, multivariable Cox proportional hazards models, Restricted Cubic Splines (RCS), and time-dependent ROC analysis, and performed subgroup analyses to assess heterogeneity. RESULTS: BMI/BRI/WC/WHtR generally displayed patterns consistent with the "obesity paradox", while ABSI did not clearly conform to this paradox. In two-piecewise Cox models, ABSI was not associated with CVM but demonstrated a significant, non-paradox association with ACM in DM at higher ABSI levels. RCS revealed nonlinear associations for BMI/BRI/WC/WHtR, while ABSI showed a nonlinear association only with ACM (inflection point: prediabetes = 0.82, DM = 0.77). In unadjusted ROC analyses, ABSI generally showed higher discriminative ability for mortality outcomes; however, after full adjustment, AUC values became similar across indicators. Subgroup analyses further indicated that the predictive performance of these indicators varied across different smoking statuses, race, and sex. DISCUSSION: The "obesity paradox" observed in BMI/BRI/WC/WHtR likely reflects their inability to distinguish fat from muscle. ABSI, which primarily captures central adiposity, did not clearly conform to this paradox; however, its incremental predictive value diminished after adjustment, suggesting substantial overlap with conventional risk factors. CONCLUSION: Unlike BMI, BRI, WC, and WHtR, ABSI did not display a clear "obesity paradox" pattern and showed higher unadjusted discrimination for mortality outcomes, although this advantage was attenuated after full adjustment, suggesting it may provide complementary risk information.